© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW
    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    2. OpenAI · Sparse Autoencoder for GPT2-Small - v5 2024
    3. GPT2-Small
    4. Residual Post - 32k
    5. 10-RES_POST_32K-OAI
    6. 29900
    Prev
    Next
    INDEX
    Explanations

    the presence of time-related phrases, particularly those beginning with "on."

    oai_token-act-pair · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 10-res_post_32k-oai
    Configuration
    jbloom/GPT2-Small-OAI-v5-32k-resid-post-SAEs/v5_32k_layer_10.pt
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    32,768
    Data Type
    torch.float32
    Hook Name
    blocks.10.hook_resid_post
    Hook Layer
    10
    Architecture
    standard
    Context Size
    64
    Dataset
    Skylion007/openwebtext
    Activation Function
    topk
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    362
    +0.10
    0.4%
    90
    +0.09
    0.4%
    269
    +0.09
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    269
    +0.10
    0.13
    90
    +0.09
    0.12
    287
    +0.09
    0.10
    Negative Logits
    ¯¯
    -0.72
    acters
    -0.69
    abwe
    -0.65
    rament
    -0.64
    EngineDebug
    -0.63
    çͰ
    -0.63
    ennes
    -0.62
    uum
    -0.60
    anan
    -0.60
    utenberg
    -0.60
    POSITIVE LOGITS
     behalf
    1.92
    erous
    1.52
     occasion
    1.32
    shore
    1.26
    etime
    1.22
    screen
    1.16
    site
    1.14
     Sundays
    1.08
     top
    1.06
     Saturdays
    1.03
    Activations Density 2.273%

    No Known Activations